| 摘要: |
| 在可控网络中,利用多agent系统是网络控制的一种重要方法.在可控网络中,多agent系统中所有agent持有的信念必须在决策前达到与网络实际状态一致,即多agent系统的信念应具有可达性,是实现网络合理控制的基础.传统的基于agent行为的信念更新模型建模过程复杂,不适合在网络控制中直接分析多agent系统信念的可达性和收敛速度.基于传统的信念更新模型,提出了信念距离的概念,并在该概念的基础上提出了新的多agent系统信念距离更新模型,并证明了该模型的合理性.该模型对多agent系统信念距离更新过程加以描述,利用线性系统对多agent系统信念收敛过程进行描述,简化了对多agent系统信念可达性和收敛速度分析的复杂性.在该模型基础上,对网络控制中多agent系统信念可达性和收敛速度进行了分析,给出了判断多agent系统信念可达性的充要条件和收敛速度的上限.另外,针对全耦合网络和无标度网络两种复杂网络的特点,分别对两种网络下多agent系统信念可达性和收敛速度进行了讨论.提出的信念距离更新模型具有良好的适应性,为判断多agent系统的信念可达性提供了有力的工具. |
| 关键词: 网络控制 多agent 信念可达性 收敛速度 可控网络 |
| DOI： |
| 分类号: |
| 基金项目:Supported by the National Natural Science Foundation of China under Grants Nos.60903161, 60903162, 90912002 (国家自然科学基金); the National Basic Research Program of China under Grant No.2010CB328104 (国家重点基础研究发展计划(973)); the China Specialized Research Fund for the Doctoal Program of Higher Education under Grant No.200802860031 (高等学校博士学科点专项科研基金); the Jiangsu Provincial Natural Science Foundation of China under Grants Nos.BK2007708, BK2008030 (江苏省自然科学基金); the Jiangsu Provincial Key Laboratory of Network and Information Security of China under Grant No.BM2003201 (江苏省网络与信息安全重点实验室资助项目); the Key Laboratory of Computer Network and Information Integration (Southeast University), Ministry of Education of China under Grant No.93K-9 (计算机网络和信息集成教育部重点实验室资助项目) |
|
| Analysis on Belief Reachability and Convergence Rate of Multi-Agent System in Controllable Networks |
|
WANG Peng,LUO Jun-Zhou,LI Wei,BIAN Zheng-Ai,QU Yan-Sheng
|
| Abstract: |
| Using a multi-agent system to control network is an important method to manipulate controllable networks. The reasonable control of network is based on the belief reachability of the multi-agent system, which means that the belief of every agent must be consistent with the real state of the network before making decisions. To research the belief reachability and convergence rate of the multi-agent system, a new model named multi-agent system belief distance updating model, which describes the updating progress of distance between agent’s belief and real network state, is proposed based on the traditional belief updating model. And the rationality of the new model is also proved. The model which transforms the belief updating progress into a linear system simplifies the analysis of belief reachability and convergence rate of the multi-agent system. Based on this model, a sufficient and necessary condition for belief reachability, and the upper limit of convergence rate of multi-agent system are proved. Besides, the belief reachability and the convergence rate of the multi-agent system in complete coupling network and scale-free network are also discussed respectively considering the characteristics of the two complicated networks. The model is adaptable to all multi-agent environments, and provides a good tool to analyze the belief reachability of the multi-agent system. |
| Key words: network control multi-agent belief reachability convergence rate controllable network |
